U.S. patent application number 10/884516 was filed with the patent office on 2006-01-05 for image reproduction.
Invention is credited to Jordi Arnabat Benedicto, Oscar Martinez, Steven John Simske, Ramon Vega.
Application Number | 20060001690 10/884516 |
Document ID | / |
Family ID | 35513385 |
Filed Date | 2006-01-05 |
United States Patent
Application |
20060001690 |
Kind Code |
A1 |
Martinez; Oscar ; et
al. |
January 5, 2006 |
Image reproduction
Abstract
A method of reproducing an image, comprising: creating a, or
using an already existing, bitmap-input image; finding zones in the
input image containing text; determining colors of pixels,
characters, or larger text items in the text zones; reproducing the
image, wherein pixels, characters or larger text items with a color
near to a primary color are reproduced in the primary color.
Inventors: |
Martinez; Oscar;
(Castelldefels, ES) ; Simske; Steven John; (Fort
Collins, CO) ; Benedicto; Jordi Arnabat; (Tarragona,
ES) ; Vega; Ramon; (Sabadell, ES) |
Correspondence
Address: |
HEWLETT PACKARD COMPANY
P O BOX 272400, 3404 E. HARMONY ROAD
INTELLECTUAL PROPERTY ADMINISTRATION
FORT COLLINS
CO
80527-2400
US
|
Family ID: |
35513385 |
Appl. No.: |
10/884516 |
Filed: |
July 2, 2004 |
Current U.S.
Class: |
347/19 |
Current CPC
Class: |
B41J 29/393
20130101 |
Class at
Publication: |
347/019 |
International
Class: |
B41J 29/393 20060101
B41J029/393 |
Claims
1. A method of reproducing an image by an ink-jet printing device,
comprising: creating a, or using an already existing, bitmap-input
image; finding zones in the input image containing text;
determining (i) colors of pixels, characters, or larger text items
in the text zones, (ii) sizes of the characters or larger text
items, (iii) a main orientation of the text in the input image;
printing the image, wherein (i) pixels, characters or larger text
items with a color near to a basic color are reproduced in the
basic color, (ii) smaller text is reproduced with a higher spatial
resolution than larger text, (iii) the image is printed in a print
direction transverse to a main reading direction of the text, based
on the main text orientation determined.
2. The method of claim 1, wherein the input image is created by
scanning or capturing a physical image and producing a bitmap
representation of it, or by converting an image represented by
structured data into the bitmap-input image.
3. The method of claim 1, wherein determining and reproducing the
color of characters or larger text items comprises: recognizing
characters by optical character recognition; averaging the colors
of the pixels associated with recognized characters or larger text
items; reproducing the characters or larger text items, when the
average color of a character or larger text item is near to a basic
color, in the basic color.
4. The method of claim 1, wherein the higher spatial resolution for
smaller text is achieved by using a higher-resolution-print mask
for the smaller text.
5. The method of claim 1, wherein, for the printing process, a page
orientation is chosen such that the print direction is transverse
to the main reading direction of the text.
6. The method of claim 1, wherein the color representation of
pixels in the input image which are to be printed in a modified
color, i.e. a basic color, is transformed into a representation of
the basic color, and the image transformed in this way is then
printed.
7. The method of claim 1, wherein pixels in the input image which
are to be printed in a modified color, i.e. a basic color, are
tagged, and wherein, during the printing process, the tagged pixels
are printed in the modified color.
8. The method of claim 1, wherein pixels associated with characters
or larger text items to be printed with a higher spatial resolution
are tagged, and wherein, during the printing process, a higher
spatial resolution is chosen for tagged pixels.
9. The method of claim 1, wherein pixels associated with text in
text zones found are tagged, and wherein the way of printing pixels
tagged as text pixels differs from that of other pixels by at least
one of: different halftone methods are applied, different spatial
or color resolutions are used, different linearization methods are
used, edges are treated in a different manner, text is underprinted
with color to increase optical density.
10. A method of reproducing an image, comprising: creating a, or
using an already existing, bitmap-input image; finding zones in the
input image containing text; determining colors of pixels,
characters, or larger text items in the text zones; reproducing the
image, wherein pixels, characters or larger text items with a color
near to a primary color are reproduced in the primary color.
11. A method of reproducing an image, comprising: creating a, or
using an already existing, bitmap-input image; finding zones in the
input image containing text; determining colors of characters or
larger text items in the text zones by recognizing characters by
optical character recognition and averaging the colors of pixels
associated with recognized characters or larger text items;
reproducing the image, wherein the characters or larger text items,
when the average color of a character or larger text item is near
to a basic color, are reproduced in the basic color.
12. A method of reproducing an image, comprising: creating a, or
using an already existing, bitmap-input image; finding zones in the
input image containing text; determining sizes of the characters or
larger text items in the text zones; reproducing the image, wherein
smaller text is reproduced with a higher spatial resolution than
larger text.
13. A method of reproducing an image by an ink-jet printing device,
comprising: creating a, or using an already existing, bitmap-input
image; finding zones in the input image containing text;
determining a main orientation of the text in the zones found in
the input image; printing the image in a print direction transverse
to a main reading direction of the text, based on the main text
orientation determined.
14. An ink-jet printing device comprising: a text finder arranged
to find text zones in a bitmap-input image; a color determiner
arranged to determine colors of pixels, characters, or larger text
items in the text zones; a size determiner arranged to determine
the size of the characters or larger text items; an orientation
determiner arranged to determine a main orientation of the text in
the input image; wherein the printing device is arranged to print
the image such that (i) pixels, characters or larger text items
with a color near to a basic color are reproduced in the basic
color, (ii) smaller text is reproduced with a higher spatial
resolution than larger text, (iii) the image is printed in a print
direction transverse to a main reading direction of the text, based
on the main text orientation determined.
15. The ink-jet printing device of claim 14, comprising a scanner
or capturing device to obtain the bitmap-input image from a
physical image.
16. The ink-jet printing device of claim 14, comprising an
image-representation converter arranged to convert an image
represented by structured data into the bitmap-input image.
17. The ink-jet printing device of claim 14, wherein the color
determiner is arranged to recognize characters by optical character
recognition, average the colors of the pixels associated with
recognized characters or larger text items; and wherein the
printing device is arranged to reproduce the characters or larger
text items, when the average color of a character or larger text
item is near to a basic color, in the basic color.
18. The ink-jet printing device of claim 14, arranged to use
higher-resolution-print masks for smaller text to achieve the
higher spatial resolution for the smaller text.
19. The ink-jet printing device of claim 14, comprising a
page-orientation turner arranged to turn the page to be printed to
an orientation in which the print direction is transverse to the
main reading direction of the text.
20. The ink-jet printing device of claim 14, comprising a color
transformer arranged to transform the color representation of
pixels in the input image which are to be printed in a modified
color, i.e. a basic color, into a representation of the basic
color.
21. The ink-jet printing device of claim 14, comprising a color
tagger arranged to tag pixels in the input image which are to be
printed in a modified color, i.e. a basic color, wherein the
printing device is arranged, during the printing process, to print
the tagged pixels in the modified color.
22. The ink-jet printing device of claim 14, comprising a
small-text tagger arranged to tag pixels associated with characters
or larger text items to be printed with a higher spatial
resolution, wherein the printing device is arranged, during the
printing process, to choose a higher spatial resolution for tagged
pixels.
23. The ink-jet printing device of claim 14, comprising a text
tagger arranged to tag pixels associated with text in text zones
found, wherein the printing device is arranged to print pixels
tagged as text pixels in a way that differs from that of other
pixels by at least one of: different halftone methods, different
spatial or color resolutions, different linearization methods,
different edge-treatment, text underprint with color to increase
optical density.
24. An image-reproduction device comprising: a text finder arranged
to find text zones in a bitmap-input image; a color determiner
arranged to determine colors of pixels, characters, or larger text
items in the text zones; wherein the image-reproduction device is
arranged to reproduce the image such that pixels, characters or
larger text items with a color near to a primary color are
reproduced in the primary color.
25. An image-reproduction device comprising: a text finder arranged
to find text zones in a bitmap-input image; a color determiner
arranged to determine colors of pixels, characters, or larger text
items in the text zones by optical character recognition and
average the colors of pixels associated with recognized characters
or larger text items; wherein the image-reproduction device is
arranged to reproduce the image such that the characters or larger
text items, when the average color of a character or larger text
item is near to a basic color, in the basic color.
26. A image-reproduction device, comprising: a text finder arranged
to find text zones in a bitmap-input image; a size determiner
arranged to determine sizes of the characters or larger text items
in the text zones; wherein the printing device is arranged to
reproduce the image such that smaller text is reproduced with a
higher spatial resolution than larger text.
27. An ink-jet printing device comprising: a text finder arranged
to find text zones in a bitmap-input image; an orientation
determiner arranged to determine a main orientation of the text in
the input image; wherein the printing device is arranged to print
the image in a print direction transverse to a main reading
direction of the text, based on the main text orientation
determined.
Description
FIELD OF THE INVENTION
[0001] The present invention relates generally to methods and
devices to reproduce an image, e.g. printing devices.
BACKGROUND OF THE INVENTION
[0002] Current techniques of manifolding and reproducing graphical
representations of information, such as text and pictures
(generally called "images") involve digital-image-data processing.
For example, a computer-controlled printing device or a computer
display prints or displays digital image data. The image data may
either be produced in digital form, or may be converted from a
representation on conventional graphic media, such as paper or
film, into digital image data, for example by means of a scanning
device. Recent copiers are combined scanners and printers, which
first scan paper-based images, convert them into digital image
representations, and print the intermediate digital image
representation on paper.
[0003] Typically, images to be reproduced may contain different
image types, such as text and pictures. It has been recognized that
the image quality of the reproduced image may be improved by a way
of processing that is specific to text or pictures. For example,
text typically contains more sharp contrasts than pictorial images,
so that an increase in resolution may improve the image quality of
text more than that of pictures.
[0004] U.S. Pat. No. 5,767,978 describes an image segmentation
system able to identify different image zones ("image classes"),
for example text zones, picture zones and graphic zones. Text zones
are identified by determining and analyzing a ratio of strong and
weak edges in a considered region in the input image. The different
image zones are then processed in different ways.
[0005] U.S. Pat. No. 6,266,439 B1 describes an image processing
apparatus and method in which the image is classified into text and
non-text areas, wherein a text area is one containing black or
nearly black text on a white or slightly colored background. The
color of pixels representing black-text components in the
black-text regions is then converted or "snapped" to full black in
order to enhance the text data.
SUMMARY OF THE INVENTION
[0006] A first aspect of the invention is directed to a method of
reproducing an image by an ink-jet printing device. The method
comprises: creating a, or using an already existing, bitmap-input
image; finding zones in the input image containing text;
determining (i) colors of pixels, characters, or larger text items
in the text zones, (ii) sizes of the characters or larger text
items, (iii) a main orientation of the text in the input image; and
printing the image, wherein (i) pixels, characters or larger text
items with a color near to a basic color are reproduced in black or
the primary color, (ii) smaller text is reproduced with a higher
spatial resolution than larger text, (iii) the image is printed in
a print direction transverse to a main reading direction of the
text, based on the main text orientation determined.
[0007] According to another aspect, a method is provided of
reproducing an image. The method comprises: creating a, or using an
already existing, bitmap-input image; finding zones in the input
image containing text; determining colors of pixels, characters, or
larger text items in the text zones; reproducing the image, wherein
pixels, characters or larger text items with a color near to a
primary color are reproduced in the primary color.
[0008] According to another aspect, a method is provided of
reproducing an image. The method comprises: creating a, or using an
already existing, bitmap-input image; finding zones in the input
image containing text; determining colors of characters or larger
text items in the text zones by recognizing characters by optical
character recognition and averaging the colors of pixels associated
with recognized characters or larger text items; reproducing the
image, wherein the characters or larger text items, when the
average color of a character or larger text item is near to a basic
color, are reproduced in the basic color.
[0009] According to another aspect, a method is provided of
reproducing an image. The method comprises: creating a, or using an
already existing, bitmap-input image; finding zones in the input
image containing text; determining sizes of the characters or
larger text items in the text zones; reproducing the image, wherein
smaller text is reproduced with a higher spatial resolution than
larger text.
[0010] According to another aspect, a method is provided of
reproducing an image by an ink-jet printing device. The method
comprises: creating a, or using an already existing, bitmap-input
image; finding zones in the input image containing text;
determining a main orientation of the text in the zones found in
the input image; printing the image in a print direction transverse
to a main reading direction of the text, based on the main text
orientation determined.
[0011] According to another aspect, an ink-jet printing device is
provided. It comprises a text finder arranged to find text zones in
a bitmap-input image; a color determiner arranged to determine
colors of pixels, characters, or larger text items in the text
zones; a size determiner arranged to determine the size of the
characters or larger text items; and an orientation determiner
arranged to determine a main orientation of the text in the input
image. The printing device is arranged to print the image such that
(i) pixels, characters or larger text items with a color near to a
basic color are reproduced in the basic color, (ii) smaller text is
reproduced with a higher spatial resolution than larger text, (iii)
the image is printed in a print direction transverse to a main
reading direction of the text, based on the main text orientation
determined.
[0012] According to another aspect, an image-reproduction device is
provided. It comprises a text finder arranged to find text-zones in
a bitmap-input image; and a color determiner arranged to determine
colors of pixels, characters, or larger text items in the text
zones. The image-reproduction device is arranged to reproduce the
image such that pixels, characters or larger text items with a
color near to a primary color are reproduced in the primary
color.
[0013] According to another aspect, an image-reproduction device is
provided. It comprises a text finder arranged to find text zones in
a bitmap-input image; and a color determiner arranged to determine
colors of pixels, characters, or larger text items in the text
zones by optical character recognition and average the colors of
pixels associated with recognized characters or larger text items.
The image-reproduction device is arranged to reproduce the image
such that the characters or larger text items, when the average
color of a character or larger text item is near to a basic color,
in the basic color.
[0014] According to another aspect, an image-reproduction device is
provided. It comprises a text finder arranged to find text zones in
a bitmap-input image; and a size determiner arranged to determine
sizes of the characters or larger text items in the text zones. The
image-reproduction device is arranged to print the image such that
smaller text is reproduced with a higher spatial resolution than
larger text.
[0015] According to another aspect, an ink-jet printing device is
provided. It comprises a text finder arranged to find text zones in
a bitmap-input image; and an orientation determiner arranged to
determine a main orientation of the text in the input image. The
printing device is arranged to print the image in a print direction
transverse to a main reading direction of the text, based on the
main text orientation determined.
[0016] Other features are inherent in the methods and products
disclosed or will become apparent to those skilled in the art from
the following detailed description of embodiments and its
accompanying drawings.
DESCRIPTION OF THE DRAWINGS
[0017] Embodiments of the invention will now be described, by way
of example, and with reference to the accompanying drawings, in
which:
[0018] FIG. 1 is a flow diagram illustrating the generation and
preparation of image data for reproduction, using three different
measures to improve image quality;
[0019] FIG. 2 is a flow diagram similar to FIG. 1 illustrating an
embodiment in which one of the measures is used, namely color
snapping;
[0020] FIG. 3 is a flow diagram illustrating color snapping in more
detail;
[0021] FIGS. 4a-b show representations of an exemplary character at
the different stages of the color-snapping procedure, wherein FIG.
4a illustrates an embodiment using color transformation, and FIG.
4b illustrates an embodiment using color tagging;
[0022] FIG. 5 is a flow diagram as FIG. 3, but including the
text-item recognition based on OCR;
[0023] FIGS. 6a-d illustrate an embodiment of the color-snapping
procedure based on OCR;
[0024] FIG. 7 is a flow diagram similar to FIG. 1 illustrating an
embodiment in which another of the measures to improve the image
quality is used, namely reproducing small characters with higher
spatial resolution;
[0025] FIG. 8 is a flow diagram which illustrates the reproduction
of small characters with higher spatial resolution in more
detail;
[0026] FIG. 9 shows an exemplary representation of characters with
different sizes reproduced with different spatial resolutions;
[0027] FIG. 10 is a flow diagram similar to FIG. 1 illustrating an
embodiment in which yet another of the measures to improve the
image quality is used, namely choosing the print direction
perpendicular to the main reading direction;
[0028] FIG. 11 is a flow diagram which illustrates printing
perpendicularly to the main reading direction in more detail;
[0029] FIGS. 12a-b illustrate that reproductions of a character may
differ when printed in different directions;
[0030] FIG. 13 is a flow diagram illustrating the reproduction of
tagged image data;
[0031] FIGS. 14a-d show components for carrying out the method of
FIG. 1 and illustrate, by exemplary alternatives, that these
components can be integrated into a single device or distributed
over several devices;
[0032] FIG. 15 is a high-level functional diagram of an image
processor;
[0033] FIG. 16 is a high-level functional diagram of a reproduction
processor.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0034] FIG. 1 is a flow diagram illustrating the generation and
preparation of image data for reproduction. Before proceeding
further with the detailed description of FIG. 1, however, a few
items of the embodiments will be discussed.
[0035] In some of the embodiments, digital image data representing
the image to be reproduced is obtained by scanning or capturing a
physical image. Scanning may be done e.g. by a scanner, and
capturing, e.g. by a video camera. A captured image may also be a
frame extracted from moving images, such as video images. A
physical image, e.g. a paper document, may be scanned and digitized
by a scanning device, which generates an unstructured digital
representation, a "bitmap", by transforming content information of
the physical image into digital data. The physical image is
discretized into small areas called "picture elements" or "pixels".
The number of pixels per inch ("ppi") in the horizontal and
vertical directions is used as a measure of the spatial resolution.
Resolution is generally expressed by two numbers, horizontal ppi
and vertical ppi; in the symmetric case, when both numbers are
equal, one number is only used. For scanners, frequently used
resolutions are 150, 300 and 600 ppi, and in the case of printing,
300, 600 and 1200 dpi are common numbers (in the case of printing,
the smallest printable unit is a "dot"; thus, rather than ppi, the
unit "dpi" (dots per inch) is often used).
[0036] The color and brightness of the paper area belonging to one
pixel is averaged, digitized and stored. It forms, together with
the digitized color and brightness data of all other pixels, the
digital bitmap data of the image to be reproduced. In the
embodiments the range of colors that can be represented (called
"color space") is built up by special colors called "primary
colors". The color and brightness information of each pixel is then
often expressed by a set of different channels, wherein each
channel only represents the brightness information of the
respective primary color. Colors different from primary colors are
represented by a composition of more than one primary color. In
some embodiments which use a cathode ray tube or a liquid crystal
display for reproduction, a color space composed of the primary
colors red, green and blue ("RGB color space") may be used, wherein
the range of brightness of each primary color, for example, extends
from a value of "0" (0% color=dark) to a value of "255" (100%
color=bright). In some systems, such as Macintosh.RTM. platforms,
this ordering may be reversed. In the example above, with values
from 0 to 255, one primary color in one pixel can be represented by
8 bits and the full color information in one pixel can be
represented by 24 bits. In other embodiments, the number of bits
used to represent the range of a color can be different from 8. For
example, nowadays scanner devices can provide 10, 12 and even more
bits per color. The bit depth (number of bits) depends on the
capability of the hardware to discretize the color signal without
introducing noise. The composition of all three primary colors in
full brightness (in the 8 bit example: 255, 255, 255) produces
"white", whereas (0, 0, 0) produces "black", which is the reason
for the RGB color space being called an "additive" color space. In
other embodiments, which use a printing device, such as an ink-jet
printer or laser printer, a "subtractive" color space is generally
used for reproduction, often composed of the primary colors cyan,
magenta and yellow. The range of each channel, for example, may
again extend from "0" (0% color=white) to "255" (100% color=full
color), able to be represented by 8 bits (as mentioned above, more
than 8 bits may be used to represent one color), but unlike the RGB
color system the absence of all three primary colors (0, 0, 0)
produces white (actually it gives the color of the substrate or
media on which the image is going to be printed, but often this
substrate is "white", i.e. there is no light absorption by the
media), whereas the highest value of all primary colors (255, 255,
255) produces black (as mentioned above, the representation may be
different on different platforms). However, due to technical
reasons the combination of all three primary colors may not lead to
full black, but a dark gray near to black. For this reason black
("Key") may be used as an additional color, the resulting color
space is then called "CMYK color space". With four colors, such as
CMYK, each represented by 8 bits, the complete color and brightness
information of one pixel is represented by 32 bits (as mentioned
above, more than 8 bits per color, i.e. more than 32 bits may be
used). Transformations between color spaces are generally possible,
but may result in color inaccuracies and, depending on the primary
colors used, may not be available for all colors which can be
represented in the initial color space. Often, printers which
reproduce images using CMY or CMYK inks are only arranged to
receive RGB input images, and are therefore sometimes called "RGB
printers". However, when colors and color spaces are discussed
herein in connection with color snapping and color reproduction,
the colors and color spaces referred are the ones actually used in
a reproduction device for the reproduction, rather than input
colors (e.g., they are CMYK in a printer with CMYK inks).
[0037] Since, in a CMYK color space, black plays a particular role
and is not a regular primary color, such as red, green, blue, or
cyan, magenta, yellow, it is often not subsumed to the "primary
colors". Therefore, the term "primary color" herein refers to one
of the regular primary colors, such as red, green, blue, or cyan,
magenta, yellow. The more generic term "basic color" is used herein
to refer to: [0038] black alone, for example, if black is the only
color, as in white-black reproduction; or [0039] one of the primary
colors and black, for example, if black is used in addition to
primary colors, as in the CMYK color space; or [0040] one of the
primary colors (without black), for example, if black is not used
in addition to primary colors, as in the RGB color space.
[0041] In some of the embodiments, the bitmap input data is not
obtained by scanning or capturing a physical image, but by
transforming an already existing digital representation. This
representation may be a structured one, e.g. a vector-graphic
representation, such as DXF, CDR, MPGL, an unstructured (bitmap)
representation, or a hybrid representation, such as CGM, WMF, PDF,
POSTSCRIPT. Creating the bitmap-input image may include
transforming structured representations into bitmap. Alternatively,
or additionally, it may also include transforming an existing
bitmap representation (e.g. an RGB representation) into another
color representation (e.g. CMYK) used in the graphical processing
described below. Other transformations may involve decreasing the
spatial or color resolution, changing the file format or the
like.
[0042] The obtained bitmap of the image to be reproduced is then
analyzed by a zoning analysis engine (i.e. a program performing a
zoning analysis) in order to distinguish text zones from non-text
zones, or, in other words, to perform a content segregation, or
segmentation. As will be explained in more detail below, the text
in the text zones found in the zoning analysis is later used in one
or more activities to improve the text image quality, such as
"color snapping", use of a font-size-dependent spatial resolution
and/or choice of a print direction transverse to a main reading
direction. Zoning analysis algorithms are known to the skilled
person, for example, from U.S. Pat. No. 5,767,978 mentioned at the
outset. For example, a zoning analysis used in some of the
embodiments identifies high-contrast regions ("strong edges"),
which are typical for text content and low-contrast regions ("weak
edges") typical for continuous-tone zones, such as pictures or
graphics. In some embodiments, the zoning analysis calculates the
ratio of strong and weak edges within a pixel region; if the ratio
is above a predefined threshold, the pixel region is considered as
a text region which may be combined with other text regions to form
a text zone. Other zoning analyses count the dark pixels or analyze
the pattern of dark and bright pixels within a pixel region in
order to identify text elements or text lines. The different types
of indication for text, such as the indicator based on strong edge
recognition and the one based on background recognition, may be
combined in the zoning analysis. As a result of the zoning
analysis, text zones are found and identified in the bitmap-input
image, e.g. as illustrated in FIG. 3 of U.S. Pat. No. 5,767,978.
Typically, but not necessarily, the zoning analysis is tuned such
that text embedded in pictures is not considered as a text zone,
but is rather assigned to the picture in which it is embedded.
[0043] In the embodiments, three different measures are applied to
improve the image quality of the text reproduced; these measures
are: (i) snapping to basic color; (ii) using higher spatial
resolution for small text; and (iii) print direction perpendicular
to the main reading direction. In some of the embodiments, only one
of the measures (i), (ii) or (iii) is used. In other embodiments,
pairs of these measures, (i) and (ii), (i) and (iii), or (ii) and
(iii) are used. Finally, in some embodiments, the combination of
all three measures, (i) and (ii) and (iii), is used.
[0044] In the framework of all three measures, optical character
recognition (OCR) may be used to identify text items (e.g.
characters) within the text zones and identify certain text-item
attributes (such as text font, text size, text orientation). In
connection with the first measure, "snapping to basic color", OCR
may also be used to determine whether individual pixels in a text
zone of the input bitmap, belong to a text item. OCR algorithms
able to identify text items and their attributes are well-known in
the art. Once a text item has been recognized by OCR, it can be
determined which pixels lie inside the recognized text item, and
which pixels lie outside; the pixels lying inside the text item are
considered as the pixels belonging to the text item (since a pixel
is an extended object, it may partly lie on the boundary of a text
item; therefore, the decision criterion may be whether the center
of a pixel lies inside or outside the text item recognized).
[0045] The first measure, "snapping to basic color" is now
explained in more detail. As already mentioned above, the terms
"color", "primary color", "color average", "color threshold", etc.,
used in this context refer to the colors, primary colors, etc.,
actually used in the reproduction device for the reproduction
(e.g., they are CMYK in a printer with CMYK inks), rather than
color in input images which may be in a different color
representation (e.g. RGB in a CMYK printer accepting RGB
input).
[0046] First, the meanings of the terms "snapping to basic color"
and "snapping to primary color" is discussed. Referring to the
above definitions of "basic color" and "primary color", the term
"snapping to basic color" includes: [0047] (a) only snapping to
black; if, although primary colors are used (as in CMYK), the
primary colors are not included in the color-snapping procedure; or
[0048] (b) only snapping to black, if black is the only color used,
as in white-black reproduction; or [0049] (c) snapping to one of
the primary colors and black, if black is used in addition to
primary colors (as in CMYK), and the primary colors are included in
the color-snapping procedure; or [0050] (d) snapping to one of the
primary colors (without black), if black is not used in addition to
primary colors (as in RGB), or if black is used, but is not
included in the color-snapping procedure.
[0051] In connection with claims 10 and 24, the term "snapping to
primary color" is used. This indicates the ability to snap to a
primary color, such as red, green, blue, or cyan, magenta, yellow,
irrespective of whether there is also a "snapping to black"; it
therefore includes the above alternatives (c) and (d), but does not
include alternatives (a) and (b).
[0052] To perform color snapping, first, the color of a pixel, or
the average color of a group of pixels forming a character or a
larger text item, such as a word, is determined. A test is then
made whether the (average) color is near a basic color, for example
by ascertaining whether the (average) color is above a basic-color
threshold, e.g. 80% black, 80% cyan, 80% magenta or 80% yellow in a
CMYK color space. If this is true for one basic color, the pixel,
or the group of pixels, is reproduced in the respective basic
color, in other words, it is "snapped" to the basic color. Such a
snapping to the basic color improves the image quality of the
reproduced text, since saturated colors rather than mixed colors
are then used to reproduce the pixel, or group of pixels.
[0053] If only one basic color is used (e.g. only black or only one
primary color), the above-mentioned threshold test is simple, since
only one basic-color threshold has then to be tested. If there are
more than one basic colors (e.g. four basic colors in a CMYK
system), it may happen that the (average) color tested exceeds two
or more of the basic color thresholds (e.g. the color has 85%
yellow and 90% magenta). In such a case, in some embodiments, the
color is then snapped to the one of the basic colors having the
highest color value in the tested color (e.g. to magenta, in the
above example). In other embodiments, no color snapping is
performed if more than one basic-color threshold is exceeded. The
basic-color threshold need not necessarily be a fixed single-color
threshold, but may combine color values of all basic colors, since
the perception of a color may depend on the color values of all
basic colors. Of course, the basic color thresholds may also depend
on the kind of reproduction and the reproduction medium.
[0054] In embodiments in which the color of pixels of a group of
pixels is averaged, and the average color is tested against the
basic-color thresholds, first a decision is taken as to which
pixels belong to the group, as already mentioned above; in some
embodiments, OCR is applied to the text zones, and the pixels
belonging, e.g. to the individual characters recognized by OCR,
form the "groups of pixel" to be averaged.
[0055] In the averaging procedure, in some embodiments, pixels of a
group having a color value considerably different from the other
pixels of the group (also called "outliers") are not included in
the average. For example, if a character is imperfectly represented
in the input-image, e.g. if a small part of a black character is
missing (which corresponds to a white spot in the case of a white
background), the character could nevertheless be correctly
recognized by the OCR, but the white pixels (forming the white
spot) are excluded from the calculation of the average color. The
exclusion of such outliers is, in some embodiments, achieved in a
two-stage averaging process in the first stage of which the
character's overall-average color is determined using all pixels
(including the not-yet-known outliers), and then the colors of the
individual pixels are tested against a
maximum-distance-from-overall-average threshold; in the subsequent
second averaging stage only those pixels are included in the
average which have a color distance smaller than this threshold,
thereby excluding the outliers. This second color average value is
then tested against the basic-color thresholds, as described above,
to ascertain whether or not the average color of the pixels of the
group is close enough to a basic color to permit their color to be
snapped to this basic color. In some of the embodiments, the
snapping thresholds mainly test hue, since saturation and intensity
will vary along the edges of the characters.
[0056] In most of the embodiments, it is not an aim of the
color-snapping procedure to improve the shape of text items of the
input image, such as imperfect characters, but only to reproduce
them as they are in a basic color, if the averaged pixel color is
close to the basic color (e.g. with regard to hue, since saturation
and intensity may vary along the edges). In other words, if a
character is imperfectly represented in the input image, e.g. if a
nearly black character has a white spot, the color-snapped
reproduced character will have the same imperfect shape (i.e. the
same white spot), but the other pixels (originally nearly black)
belonging to the character will be reproduced in full black (of
course, this is only exemplary, since the "black color" and "white
color" can be other background or text hues, dependent on the
histogram of the "text and background" areas in a particular case).
In some of the embodiments, this is achieved by not modifying the
color of outliers; the definition which defines which pixels are
outliers may be the same as the one described above in connection
with the exclusion of outlier pixels from the averaging procedure,
or may be another independent definition (it may, e.g. use another
threshold than the above-mentioned
maximum-distance-from-overall-average threshold).
[0057] However, in some of the embodiments, color snapping may be
combined with a "repair functionality" according to which all
pixels of a character--including outliers, such as white spots--are
set to a basic color, if the average color of the character
(including or excluding the outliers) is close to the basic color.
In such embodiments, not only the color, but also the shape or the
characters to be reproduced is modified.
[0058] There are different alternative ways in which color snapping
is actually achieved in the "reproduction pipeline" (or "printing
pipeline", if the image is printed). For example, the printing
pipeline starts by creating, or receiving, the bitmap-input image,
and ends by actually printing the output image.
[0059] In some of the embodiments, the original color values in the
bitmap-input image of the pixels concerned are replaced (i.e.
over-written) by other color values representing the basic color to
which the original color of the pixels is snapped. In other words,
the original bitmap-input image is replaced by a (partially)
modified bitmap-input image. This modified bitmap-input image is
then processed through the reproduction pipeline and reproduced
(e.g. printed) in a usual manner.
[0060] In other embodiments, rather than replacing the original
bitmap-input image by its color-snapped version, the original image
data is retained unchanged and the snapping information is added to
the bitmap-input image. The added data is also called a "tag", and
the process of adding data to a bitmap image is called "tagging".
Each pixel of the bitmap-input image may be tagged, for example, by
providing one additional bit per pixel. A bit value of "1", e.g.
may stand for "to be snapped to basic color" and a bit value of "0"
may stand for "not to be snapped", in the case of only one basic
color. More than one additional bit may be necessary if more than
one basic color is used (e.g. 0="not to be snapped", 1="to be
snapped to black", 2="to be snapped to first primary color", 3="to
be snapped to second primary color", etc.); this is also called
palette or lookup-table (LUT) snapping. In embodiments using
tagging the actual "snapping to basic color" is then performed at a
later stage in the reproduction pipeline, for example, when the
bitmap-input image is transformed, using a color map, into a print
map which represents the amounts of ink of different colors applied
to the individual pixels (or dots).
[0061] The second measure to improve the image quality of text
(measure (ii)) is to reproduce smaller text (e.g. characters of a
smaller font size) with a higher spatial resolution than larger
text (e.g. characters of a larger font). Generally, the number of
different reproducible colors (i.e. the color resolution) and the
spatial resolution are complementary quantities: If, on the one
hand, the maximum possible spatial resolution is chosen in a given
reproduction device (e.g. an ink-jet printer), no halftoning is
possible so that only a small number of colors can be reproduced
(or, analogously, in white-black reproduction, only white or black,
but no gray tones can be reproduced). On the other hand, if a lower
spatial resolution is chosen, a larger number of colors (in
white-black reproduction: a number of gray tones) may be
reproduced, e.g. by using halftone masks.
[0062] Generally, there are different spatial resolutions in a
printing device: (i) printing resolution, (ii) pixel size, and (ii)
halftoning resolution. [0063] the printing resolution is given by
the number of dots that can be reproduced in a certain distance;
for example, in an ink-jet printer it is given by the number of
drops that can be fired in a certain distance. The printing
resolution can be relatively high, 4800 dpi, for instance; [0064]
the pixel size is the size of the discretized cells used in the
data representation of the image to be printed; it can be
relatively small; for instance, the pixel size may be equal to dot
size, allowing a 4800 ppi resolution in the example above; [0065]
the halftoning resolution corresponds to the size of the halftoning
window, or cell. A halftoning window normally includes a plurality
of pixels to allow mixing of colors. For example, the halftoning
window can extend over 32, 64, 128, 256, etc. pixels. The bigger
the halftoning window, the bigger are the possibilities of mixing
colors, i.e. the better is the color resolution, and the smaller is
number of lines per inch that can be reproduced, i.e. the smaller
the effective spatial resolution. The halftoning resolution defines
the spatial resolution of the reproduced image. Thus, color
resolution and spatial resolution are complementary.
[0066] It has been recognized that an improved perceived text image
quality can be achieved by using a better color resolution in
larger text fonts and a better spatial resolution in smaller text
fonts. Therefore, according to measure (ii), the sizes of
characters or larger text items (such as words) in the text zones
are determined, and smaller text is then reproduced (e.g. printed)
with a higher spatial resolution than larger text.
[0067] In some of the embodiments, the determination of the
characters or larger text items is based on OCR; typically, OCR not
only recognizes character, but also provides the font sizes of
recognized characters.
[0068] The reproduction of characters and smaller text items with
higher spatial resolution is, in some of the embodiments, achieved
by using a higher-resolution print mask for smaller text.
"Higher-resolution print mask", of course, does not necessarily
mean the above-mentioned extreme case of an absence of halftoning;
it rather means that the size of the halftoning window is smaller
than in a lower-resolution print mask, but if the window size is
not yet at the minimum value (which corresponds to the pixel size),
there may still be some (i.e. a reduced amount of) halftoning. In
some embodiments, if both smaller and larger characters are found
in a text zone, a sort of hybrid print mask is used in which
regions forming higher-resolution print masks (i.e. regions with
bigger halftoning windows) are combined with regions forming
lower-resolution print masks (i.e. regions with smaller halftoning
windows).
[0069] In some of the embodiments, the printing resolution can be
changed "on the fly", i.e. during a print job. In such embodiments,
the trade-off between image quality and throughput may be improved
by choosing a smaller printing resolution when small fonts are to
be printed, rather than a smaller halftoning window. For in a
typical scanning printer, more passes have to be made to increase
paper axis resolution, or in a page-wide-array system the advance
speed is lowered, when a smaller printing resolution is used. Some
embodiments can print both at low and high print resolution grids;
in these embodiments, a higher-print-resolution grid is used in
regions with small text items (resulting in a higher number of
passes in a scanning printing system, or a lower advance speed in a
page-wide array system), but printing with a lower-print-resolution
grid is resumed in regions without small text items (resulting in a
smaller number of passes in a scanning printing system, or a higher
advance speed in a page-wide array system). As a result, throughput
is increased, while good image quality is maintained.
[0070] Normally, reproducing smaller text with higher spatial
resolution requires the input-image information to be available
with a sufficiently high spatial resolution. However, it is not
necessary for this information to be a priori available. Rather, in
some embodiments, the bitmap-input image is, at a first stage, only
created (e.g. scanned) with a smaller spatial resolution. If it
then turns out, after text-zone finding and OCR have been
performed, that a higher-resolution input bitmap is required due to
the presence of small-font text, another scan of the image to be
reproduced is performed, now with the required higher spatial
resolution.
[0071] Typically, print masks are not used at the beginning of the
printing pipeline to modify the bitmap-input image, but rather
later in the pipeline, when the print map representing the amounts
of ink to be applied to pixels (or dots) is generated. Therefore,
in some of the embodiments, the bitmap-input image is not modified
in connection with the different spatial resolutions with which it
is to be reproduced, but it is tagged. In other words, data is
added to the bitmap-input image indicating which regions of the
image are to be reproduced with which resolutions. The regions may,
e.g. be characterized by specifying boundaries of them, or by
tagging all pixels within a region with a value representing the
respective spatial resolution.
[0072] The third measure to improve the image quality of text
(measure (iii)) is to choose the print direction transverse
(perpendicular) to the main human-reading direction. This measure
is useful when an ink-jet printer is used, for example The print
direction is the relative direction between the ink-jet print head
and the media (e.g. paper) onto which the ink is applied; in the
case of a swath printer with a reciprocating print head it is
typically transverse to the media-advance direction, but in the
case of a page-width printer it is typically parallel to the
media-advance direction.
[0073] It has been recognized that most users prefer or find value
in printing perpendicular to the reading direction because: [0074]
(i) the vertical lines of most letters (which are, on average,
longer and more straight than the horizontal lines) mask typical
ink-jet defects and artifacts due to spray, misdirected ink drops,
etc. For example, if the spray provoked by ink-drops tails fall on,
or under, a fully inked area; the artifact tails are not visible;
this will happen more frequently with a print direction
perpendicular to the reading direction (in other words, the
"visible drops tails vs. character type or size ratio" is smaller
with a print direction perpendicular to the reading direction);
[0075] (ii) the human reader pays less attention to the vertical
direction of a document (perpendicular to the reading direction)
than to the horizontal direction. Defects in the document's
vertical direction are normally less annoying for human readers.
Besides, if the ink-drops tails are so "long" that they merge among
characters, this would affect a lot the reading clarity of a text.
Since, in the vertical direction, the space between characters (the
line space) is bigger than in the horizontal direction, this
merging effect is lower in the vertical direction. Thus, the
reading clarity is not so much affected due to the merging effect
with a "vertical" print direction, i.e. the print direction
perpendicular to the reading direction.
[0076] Thus, the human reader is less sensitive to
character-reproduction defects at those parts of the characters
which are transverse to the reading direction than those which are
parallel to it. For example, if a "T" is considered, a defect at
the vertical edge of the T's vertical bar would be less annoying
than a defect at the horizontal edge of the T's horizontal bar.
Accordingly, the perceived image quality of text can be improved by
choosing the printing direction perpendicular to the reading
direction.
[0077] Since a whole page is normally printed using the same print
direction, a compromise is made when a page contains text with
mixed orientations, e.g. vertically and horizontally oriented
characters (wherein "vertical character-orientation" refers to the
orientation in which a character is normally viewed, and
"horizontal character-orientation" is rotated by 90.degree. to it).
In the Roman, and many other alphabets, the reading direction is
transverse to the character orientation, i.e. it is horizontal for
vertically-oriented characters and vertical for
horizontally-oriented characters. Then, the main reading direction
of the text on this page is determined, e.g. by counting the
numbers of horizontally and vertically-oriented characters in the
text zones of the page and considering the reading direction of the
majority of the characters as the "main reading direction". Other
criteria, such as text size, font, etc. may also be used to
determine the main reading direction. For example, a different
weight may be given to characters of different fonts, since the
sensitivity to these defects may be font-dependent; e.g., a
sans-serif, blockish font like Arial will produce a greater
sensitivity to these defects than a serif, flowing font such as
Monotype Corsiva. Consequently, a greater weight may be assigned to
Arial characters than Monotype Corsiva characters, when the
characters with horizontal and vertical orientations are counted
and the main reading direction is determined. The orientation of
the characters can be determined by OCR. The print direction is
then chosen perpendicular to the main reading direction.
[0078] The main reading direction may vary from page to page since,
for example, one page may bear a majority of vertically oriented
characters, and another page a majority of horizontally oriented
characters. In the embodiments, each page of the bitmap-input image
is tagged with the one-bit tag indicating whether the main reading
direction of this page is horizontal or vertical. This
reading-direction tag is then used in the printing pipeline to
assure that the main reading direction is chosen perpendicular to
the print direction. In most printers, the print direction is
determined by the structure of the print heads and the
paper-advance mechanism, and cannot be changed. Therefore, the
desired relative orientation between the main reading direction of
the image to be printed and the print direction can be achieved by
virtually rotating the bitmap-input image or the print map
representing the amounts of ink to be printed. If the
reading-direction tag for a certain page indicates that the
orientation of the main reading direction of the bitmap-input image
data is transverse to the print direction, no such virtual rotation
is performed. By contrast, if the reading-direction tag indicates
that the main reading direction of the image data is parallel to
the print direction, a 90.degree. rotation of the image data is
performed. The subsequently printed page therefore has the desired
orientation.
[0079] Of course, the print media is provided in such a manner that
both orientations can alternatively be printed. In some of the
embodiments, the format of the print media used (e.g. paper) is
large enough to accommodate both portrait and landscape orientation
(for example, a DIN A4 image may alternatively be printed on a DIN
A3 paper sheet in portrait or landscape format, as required). In
other embodiments, the image size may correspond to the print media
size (e.g. DIN A4 image size and DIN A4 print-media size), and the
printing device has at least two different paper trays, one
equipped with paper in the portrait orientation, the other one in
the landscape orientation. In these embodiments, the printing
device is arranged to automatically supply a portrait-oriented
paper sheet if the page is printed in portrait orientation, and a
landscape-oriented paper sheet if it is printed in landscape
orientation. Thus, the reading-direction tag not only controls
whether the image data are virtually rotated by 90.degree., but
also whether portrait-oriented or landscape-oriented paper is used
for printing the tagged page.
[0080] Generally, there is a trade-off between image quality (IQ)
and throughput (mainly print speed). Depending on the printing
system, such as page-wide-array printing systems, scanning-printing
systems, etc., the page orientation influences the print speed. For
instance, in a page-wide-array system, landscape orientation could
typically be printed faster than portrait, for instance. In some
embodiments, the printing device enables the final user to select a
"fast print mode" (without using the automatic selection of a
transverse print direction, described above, but always using a
high-throughput direction, such as landscape) or a "high IQ print
mode" (with such an automatic choice).
[0081] In some of the embodiments, further measures are applied to
improve the image quality of reproduced text: in the text zones
found, halftone methods, print masks, resolutions and/or edge
treatments may be applied which are different from those used in
the picture zones or other non-text zones. Furthermore, text may be
underprinted with color to increase the optical density (needing
then less number of print passes to achieve the same perceived
optical density). In order to achieve such a different treatment of
text and picture, pixels or regions of pixels associated with text
in text zones found are tagged such that the tagging indicates that
the text-particular halftone methods, resolutions, linearization
methods, edge treatments and/or text underprintings are to be
applied to the tagged pixels or regions of pixels.
[0082] The third measure to improve the image quality of text
(choosing print direction transverse to main reading direction) is
an ink-jet-specific measure; it will therefore be used in
connection with ink-jet printing, and the embodiments of
reproducing devices implementing the third measure are ink-jet
printing devices. The first measure (snapping to black and/or
primary color) and the second measure to improve image quality
(reproducing smaller text with a higher spatial resolution than
larger text) are not only useful for ink-jet printing, but also for
other printing technologies, such as electrostatic-laser printing
and liquid electrophotographic printing, and, furthermore, for any
kind of color reproduction, including displaying the image in a
volatile manner on a display, e.g. on a liquid-crystal display or a
cathode-ray tube. The three measures may be implemented in the
reproduction device itself, i.e. in an ink-jet printing device, a
laser printing device or a computer display, or in an image
recording system, such as a scanner (or in a combined image
recording and reproducing device, such as a copier). Alternatively,
the methods may be implemented as a computer program hosted in a
multi-purpose computer which is used to transform or tag bitmap
images in the manner described above.
[0083] Returning now to FIG. 1, it shows a flow diagram
illustrating the process of generating and preparing image data for
reproduction using three different measures to improve image
quality. If no digital-data representation of the original image is
available, the original image, e.g. a sheet of paper with the image
printed on it is scanned, and a digital bitmap representation of it
is generated at 10. Alternatively, if a structured digital-data
representation of the image to be reproduced is available, e.g. a
vector-graphics image, it is transformed into a bitmap
representation at 20. In the bitmap obtained, the image is
rasterized in a limited palette of pixels, wherein the color of
each pixel is typically represented by three or four color values
of the color space used, e.g. the RGB or CMYK values. In still
further cases, a bitmap representation of the image to be
reproduced may already be available, but the available
representation may not be appropriate; e.g. the colors may be
represented in a color space not used here; rather than generating
a bitmap or transforming structured image data into a bitmap, the
existing bitmap representation is then transformed into an
appropriate bitmap representation, e.g. by transforming the
existing color representation (e.g. RGB) into another color
representation (e.g. CMYK).
[0084] At 30, the bitmap is used as an input image for further
processing. At 35, a zoning analysis is performed on the
bitmap-input image to identify text zones, e.g. as illustrated in
FIG. 3 of U.S. Pat. No. 5,767,978. At 40, the input image is
prepared for reproduction with improved image quality of text in
the text zones. As a first measure, at 41, the color of text items,
e.g. characters is determined and snapped to one of the primary
colors and black, if the original color of the character is near to
the primary color or black. The snapping to primary color or black
may either be effected by transforming the color of the pixels
belonging to the character in the bitmap-input image, or by tagging
the respective pixels of the image. As a second measure, at 42, the
sizes of the characters in the text zones are determined, and the
bit regions representing small characters are tagged so that the
small characters are reproduced with a higher spatial resolution.
As a third measure, at 43, the main orientation of the text in the
page considered is detected, and the main reading direction is
concluded from it. The page is then tagged so that it is reproduced
with the print direction perpendicular to the main reading
direction. Finally, at 50, the image is printed with the snapped
colors, higher spatial resolution for small characters and a print
direction perpendicular to the main reading direction.
[0085] Whereas FIG. 1 shows the three measures to improve image
quality 41, 42 and 43 in combination, FIGS. 2, 7 and 10 illustrate
other embodiments in which only one of the measures 41, 42 or 43 is
used. There are still further embodiments which combine measures 41
and 42, 41 and 43 and 42 and 43, respectively. The remaining
figures illustrate features of the measures 41, 42, 43, and
therefore refer both to the "combined embodiment" of FIG. 1 and the
"non-combined embodiments of FIGS. 2, 7 and 10.
[0086] FIG. 2 is a flow diagram similar to FIG. 1 illustrating an
embodiment in which only one of the measures of FIG. 1 is
performed, namely measure 41, "snapping to primary color or black".
Therefore, measures 42 and 43 are not present in FIG. 2. Since
color snapping to primary color or black is not only useful in
printing, but also when images are reproduced on video screens,
etc., reproducing the image at 50 does not refer specifically to
printing. Apart from these differences, the embodiment of FIG. 2
corresponds to FIG. 1.
[0087] FIG. 3 is a flow diagram illustrating the color-snapping
procedure (box 41 of FIGS. 1 and 2) in more detail for an
individual text item in a text zone. First, at 411, those pixels
are detected which belong to the text item (e.g. character)
considered. This detection may be based on OCR (FIG. 5) or on a
different method; for example, a cluster-detection method which
considers a cluster of similarly colored pixels as a "text item".
Then, at 412, the average color of the text item's pixels is
determined (the average may, for example, the mean, or the median,
depending on the print technology and application). As described
above, in some of the embodiments pixels having a color far away
from the text item's average color are not included in the
averaging procedure and, therefore, do not influence the average
determined at 412 (as described, this can be achieved by a
two-stage averaging procedure, in the first stage of which the
pixels with a color far away from the average color are determined
and excluded, and in the second stage of which the final color
average, not using those pixels, is determined). At 413, it is
ascertained whether the text item's average color is near a primary
color or black. If this is true, the pixels belonging to the text
item are transformed to the primary color or black, or are tagged
so that they are reproduced in the primary color or black later in
the reproduction pipeline. As explained above, in some of the
embodiments those of the text item's pixels having a color far away
from the text item's color average are not snapped to the primary
color or black, in order not to change the text item's shape, but
rather limit the effect of the color-snapping procedure to an
improvement of the text item's color reproduction.
[0088] FIGS. 4a and 4b show representations of an exemplary
character, an "H", at the different stages of the color-snapping
procedure, wherein FIG. 4a illustrates an embodiment using color
transformation, and FIG. 4b illustrates an embodiment using color
tagging. A cutout with the original bitmap-representation of the
character considered is shown at the left-hand side of FIGS. 4a and
4b. Pixels to which the color "white" is assigned, are reproduced
in white, pixels in a primary color (e.g. magenta) or black are
shown in black, and pixels which have a color near to the primary
color (e.g. magenta) or black are hatched. As can be seen, some of
the character's pixels in the original bitmap-representation are in
the primary color, or black, whereas others are near to the primary
color, or black. This may, for example, be a scanning artifact:
Assume that, in an original paper document, the character
considered here was printed in a primary color (e.g. magenta) or
black. Typically, at some of the pixels, the scanner did not
recognize the primary color, or black, but rather recognized a
slightly different color near to the primary color, or black.
[0089] During the averaging procedure described above, it is then
determined that the average color of the character considered is
near to the primary color (e.g. magenta) or black. In the
embodiment according to FIG. 4a, the pixels belonging to the
character are then transformed in the original
bitmap-representation to the primary color (e.g. magenta), or
black. This is illustrated by the bitmap representation shown in
the middle of FIG. 4a. In other words, the original bitmap-input
image is replaced by a modified one. Then, the character is
reproduced, e.g. printed or displayed, according to the modified
bitmap-representation, as illustrated at the right-hand side of
FIG. 4a.
[0090] According to another embodiment illustrated by FIG. 4b, the
character's original bitmap-representation is not replaced by a
modified one, but rather the original representation is tagged by
data specifying which pixels are to be reproduced in the primary
color (e.g. magenta) or black. In the example shown in FIG. 4b, all
pixels to be reproduced in the primary color (e.g. magenta), or
black, are tagged with a "1", whereas the remaining bits are tagged
with a "0". Of course, in other embodiments tags with more than one
bit are used to enable snapping to more than one color, e.g. to
three primary colors and black, to be tagged. In the example of
FIG. 4b, also those pixels are tagged with "1" which already in the
original bitmap representation are represented in the primary
color, or black. This, of course, is redundant and may be omitted
in other embodiments.
[0091] In the reproduction pipeline, the tags indicate that the
tagged pixels are to be printed in the primary color, or black,
although the color assigned to the respective pixel in the bitmap
representation indicates a different color. Finally, the character
is reproduced in the primary color, or black, as shown at the
right-hand side of FIG. 4b. Although the internal mechanism of
"color snapping" is different in FIGS. 4a and 4b, the reproduced
representations are identical.
[0092] FIG. 5 is a flow diagram similar to FIG. 3, but is more
specific in showing that the detection of pixels belonging to a
text item is performed using optical character recognition (OCR).
The recognized text items are therefore characters. In principle,
OCR recognizes characters by comparing patterns of pixels with
expected pixel patterns for the different characters in different
fonts, sizes, etc., and assigns that character to the pixel pattern
observed whose expected pixel pattern comes closest to the pixel
pattern observed. As a by-product, OCR is able to indicate which
pixels belong to the character recognized, and which pixels are
part of the background.
[0093] FIGS. 6a to 6d illustrate an OCR-based embodiment of the
color-snapping procedure in more detail. Similar to the example of
FIG. 4, a cut-out of a bitmap-input image is shown in FIG. 6a, now
carrying a representation of the character "h". Unlike FIG. 4, the
"h" not only has primary-color pixels (or black pixels) and
near-primary color pixels (or near-black pixels), but also has some
white spots. Furthermore, there is some colored background, i.e.
some isolated colored pixels around the character "h" (those
background pixels may have a primary color, or black, or any other
color).
[0094] After having applied OCR to this exemplary bitmap, it is
assumed that the OCR has recognized the character "h". In the
subsequent FIG. 6b, the bitmap-representation has not been changed,
but only the contour of the recognized "h" has been overlaid with
the recognized character's contour. This illustrates that the
process has awareness of which pixels belong to the character
recognized, and which ones do not. As can be seen, due to the
discrete character of the bitmap and the size of the individual
pixels, some of the pixels at the character's contour are partially
within the character's contour and, to some extent, outside it. A
pixel may not be considered as a pixel belonging to the character,
if, for instance, its center is located outside the character's
contour.
[0095] In the subsequent color-averaging procedure all pixels
belonging to the recognized character, according to the above
definition, are included, except those pixels having a color far
away from the average. In other words, the white spots are not
included. Provided that the color average determined in this manner
is near to a primary color (e.g. magenta), or black, the color of
the pixels belonging to the character recognized is then snapped to
the primary color (e.g. magenta), or black, except those pixels
which initially had a color far away from the average color, i.e.
except the white spots.
[0096] The result is illustrated in FIG. 6c, still together with
the contour of the character recognized: Originally primary-color
pixels (or black pixels) belonging to the character remain
primary-color pixels (or black pixels); originally
near-primary-color pixels (or near-black pixels) belonging to the
character are snapped to the respective primary color (or black);
pixels with a color far from a primary color (or far from black)
belonging to the character remain unchanged; and pixels which do
not belong to the character remain unchanged, too.
[0097] Finally, FIG. 6d shows the character without the character's
contour, i.e. it shows the character as it is reproduced, for
example printed on a print media. As can be seen, neither the shape
of the character nor the background has been modified, but only the
character's color representation has been improved. A reason for
not modifying the character's shape and the background is
robustness against OCR errors: If, for example, a cluster of
colored pixels in the input-bitmap image is similar to two
different characters, and the OCR recognizes the "wrong" one, this
error will only influence the color of the reproduced cluster, but
not its shape, thereby leaving a chance for the human reader to
perceive the correct character.
[0098] FIG. 7 is a flow diagram similar to FIGS. 1 and 2
illustrating an embodiment in which only another of the measures of
FIG. 1 is performed, namely measure 42, "reproducing small
characters with higher spatial resolution". Therefore, the measures
41 and 43 are not present in FIG. 7. Since reproducing small
characters with higher spatial resolution is not only useful in
printing, but also when images are reproduced on video screens,
etc., reproducing the image at 50 does not refer specifically to
printing. Apart from these differences, the embodiment of FIG. 7
corresponds to FIG. 1.
[0099] FIG. 8 is a flow diagram which illustrates the procedure of
reproducing small characters with higher spatial resolution (box 42
of FIGS. 1 and 7) in more detail. A text-size threshold below which
text is reproduced with higher spatial resolution is used, as
indicated at 421. For example, the threshold may specify a certain
font size, so that all characters with a font size below the
threshold are reproduced with a higher spatial resolution than the
other, larger characters, which, due to the complementarity between
spatial and color resolution, are printed with a higher color
resolution. The impact on image quality of the spatial resolution
chosen may have significant font dependencies; for instance, Arial
will be affected more than Times Roman, etc. Thus, in some
embodiments, different font-size threshold specific for different
fonts (e.g. Arial and Times Roman) are used. At 422, text items
(e.g. characters) within the text zones are recognized by OCR. As a
by-product of the OCR, the size of the text items (e.g. the font
size of characters) is detected, as indicated at 423. If the text
item's size is below the text-size threshold, the pixels of the
text item, or a pixel region including the text item, are, or is,
tagged at 423, so that it can be reproduced with a higher spatial
resolution than the spatial resolution used for larger text items
above the threshold. Incidentally, such a distinction by text size
may decrease throughput; thus, in some embodiments, it is only made
when selected by the final user, if throughput demands warrant such
a distinction by text size.
[0100] FIG. 9 illustrates what results are achieved when characters
having different sizes, here the characters "H.sub.2O", are
reproduced with different spatial resolutions. By applying OCR to
the bitmap-input image, the characters "H", "2" and "O" are
recognized (box 422 of FIG. 8). As a by-product of the OCR, the
font sizes of these characters are also detected; in the example
shown in FIG. 9 the font size of "2" is only about half of the font
size of "H" and "O" (box 423 in FIG. 8). Assuming that the smaller
font size is below the threshold (box 421 of FIG. 8), a region
including the "2" in the bitmap-input image is then tagged to
indicate that this region is to be reproduced with a higher spatial
resolution than the other regions. As a consequence, in some of the
embodiments (e.g. embodiments which always use the highest printing
resolution), at the end of the reproduction pipeline a print mask
is chosen for the tagged region which has a smaller halftoning
window, whereas the other regions are reproduced using a print mask
with a larger halftoning window. For example, the smaller
halftoning-window size corresponds to a resolution of 600 ppi,
whereas the larger halftoning-window size corresponds to a
resolution of 300 ppi. FIG. 9 shows a grid with the different
halftoning-window sizes, the characters "H.sub.2O" as they are
actually printed, and contour lines indicating how these characters
would appear when reproduced with a perfect spatial resolution. As
can be seen in FIG. 9, due to the discrete nature of the
halftoning-window, the shapes of the characters actually printed
differ from the ideal shapes; as can further be seen, this
difference, in absolute terms, is smaller for the smaller character
"2" than for the larger characters "H" and "O". On the other hand,
since the larger halftoning-window provide a higher color
resolution, the colors of the larger characters "H" and "O" can
generally be reproduced with a better quality than the color of the
smaller character "2".
[0101] In other embodiments, the printing resolution can be changed
"on the fly", i.e. during a print job. In such embodiments, the
trade-off between image quality and throughput may be improved by
choosing a smaller printing resolution when small fonts are to be
printed, rather than a smaller halftoning window. For in a typical
scanning printer, more passes have to be made to increase paper
axis resolution, or in a page-wide array system the advance speed
is lowered, when a smaller printing resolution is used. Some
embodiments can print both at low and high print resolution grids;
in these embodiments, a higher-print-resolution grid is used in
regions with small text items (resulting in a higher number of
passes in a scanning printing system, or a lower advance speed in a
page-wide array system), but printing with a lower-print-resolution
grid is resumed in regions without small text items (resulting in a
smaller number of passes in a scanning printing system, or a higher
advance speed in a page-wide array system). As a result, throughput
is increased, while good image quality is maintained.
[0102] FIG. 10 is a flow diagram similar to FIGS. 1, 2 and 7
illustrating an embodiment in which the third of the measures of
FIG. 1 is performed without the others, namely measure 43,
"choosing the print direction perpendicular to the main reading
direction". Apart from the fact that the other measures 41 and 42
are not present in FIG. 10, it corresponds to FIG. 1.
[0103] FIG. 11 is a flow diagram which illustrates the procedure of
choosing the print direction perpendicular to the main reading
direction (box 42 of FIGS. 1 and 10) in more detail. At 431, the
orientations of the text items (e.g. characters) in the text zones
of the page considered are determined. For example, this can be
achieved by applying OCR to the text, since this provides, as a
by-product, the orientations of the characters recognized. At 432,
the main reading direction of text in the page considered is
determined. For example, the orientation of the majority of
characters in the page considered is taken as the main orientation
of text. If the reading direction is perpendicular to the character
orientation (as is the case in the Roman alphabet), the main
reading direction of the text is determined to be perpendicular to
the main orientation of the text. For example, in a page in which
the majority of characters are vertically oriented, the main text
orientation is vertical, and the main reading direction is
horizontal. At 433, the page is then tagged to indicate the
direction in which it is to be printed. For example, a tag bit "1"
may indicate that the virtual image to be printed has to be turned
by 90.degree. before it is printed, whereas the tag bit "0" may
indicate that the virtual page need not be turned. As already
mentioned above, there is a trade-off between image quality (IQ)
and throughput. For instance, in a page-wide-array system,
landscape orientation could typically be printed faster than
portrait, for instance. In some embodiments, the printing device
enables the final user to select a "fast print mode" (without using
the automatic selection of a transverse print direction, described
above, but always using a high-throughput direction, such as
landscape) or a "high IQ print mode" (with such an automatic
choice).
[0104] FIG. 12 illustrates what a reproduced character may look
like when printed parallel (FIG. 12a) and perpendicular (FIG. 12b)
to the main reading direction. In both cases shown, the orientation
of the exemplary character "h" is vertical. Consequently, the
reading direction is horizontal. As is drawn in FIG. 12 in an
exaggerated manner, the actual reproduction of the character is not
perfect, but some ink will inevitably be applied to the white
background outside the character's contour. This effect is
typically more pronounced in the print direction than transverse to
it, as a comparison of FIGS. 12a and 12b illustrates. The perceived
image quality is better in the case of FIG. 12b, in which the print
direction is perpendicular to the reading direction. By the measure
described in connection with FIG. 11, the major part of the text in
a page is printed as in FIG. 12b, whereby the overall text image
quality is improved.
[0105] FIG. 13 illustrates how tagged data are reproduced; in other
words, it illustrates box 50 of FIGS. 1, 2, 7 and 10 in more
detail. For simplicity, three different activities, 51, 52, 53,
pertaining to the treatment of tagged image data are shown in a
combined manner in FIG. 13. Of course, FIG. 13 is also intended to
illustrate those embodiments in which only the activities 51, 52 or
53, or pairs, such as 51 and 52, 51 and 53 or 52 and 52, are
performed.
[0106] If an image is to be reproduced, it is ascertained whether
tags are assigned to the image which have to be taken into account
in the reproduction procedure. At 51, it is ascertained whether
pixels of the image or regions of pixels carry color-snapping tags
indicating that the respective pixels are to be reproduced in a
primary color or black. If such a tag is found, the respective
pixel is reproduced in the primary color, or black, indicated by
the tag. Thereby, the color of the pixels which is still indicated
in the bitmap is effectively "overridden".
[0107] At 52, it is ascertained if pixels or pixel regions are
tagged to be reproduced with a higher spatial resolution. For the
pixels or pixel regions tagged in this manner, a high-resolution
mask is used for the subsequent reproduction of the image (or the
printer is switched to a higher-printing-resolution grid, if
applicable). At 53, it is ascertained whether a page to be printed
is tagged with regard to the print direction. If a tag is found
indicating that, with the present orientation of the virtual image
in memory, the image would not be printed in the desired print
direction, the virtual image is rotated so that it is printed with
a print direction perpendicular to the main reading direction.
Finally, at 54, the image is actually displayed or printed, in the
described manner directed by the tags in 51, 52 and/or 53.
[0108] FIG. 14a to FIG. 14d show components for carrying out the
method of FIG. 1 and illustrate, by four exemplary alternatives,
that these components can be integrated into a single device or
distributed over several devices.
[0109] FIG. 14a illustrates a copier 1000, which is, e.g., an
ink-jet color copier. It has a scanning part 1003 with a scan bed
1002 which can be covered by a scan lid 1001. The scanning part
1003 is able to scan colored images printed on a printing media,
e.g. a paper sheet, and to generate a digital representation of the
original printed image, the bitmap-input image. In order to be able
to reproduce images already existing in a digital representation,
the copier 103 may also have a memory 1004 for storing digital
images. An image processor 1005 is arranged to receive the
bitmap-input images to be reproduced, either from the scanning part
1003 or the memory 1004. It processes these images, for example by
transforming near-primary colors (and near-black) to primary colors
(and black) and/or adds tags relating to color snapping, spatial
resolution and/or print direction, as explained above. A printing
unit 1006 including a print processor 1007 is arranged to produce
the print out of the image from the image processor 1005 on a print
media, e.g. a paper sheet 1008. The printer 1006 may have two paper
trays 1009 and 1010, as shown in FIG. 14a. The print processor 1007
follows the instructions represented by the tags, e.g. causes the
use of a primary color or black instead of the color represented in
the bitmap, causes the use of a high-resolution print mask for
tagged pixel regions and/or causes a page to be printed in the
direction perpendicular to the main reading direction, according to
the printing-direction tag, and finally produces a map representing
the amounts of ink of the different available colors to be applied
to the different raster points on the print media 1008. Ink-jet
print heads comprised in the printing unit 1006 finally apply the
inks according to this and produce the final print-out on the paper
sheet 1008.
[0110] The copier 1003 has two paper trays, 1009 and 1010; for
example, paper tray 1009 contains paper in portrait orientation,
and paper tray 1010 contains paper in landscape orientation. The
print processor 1007 is also coupled with a paper-tray-selection
mechanism such that, depending on the printing-direction tag, pages
to be printed in portrait orientation are printed on
portrait-oriented paper, and pages to be printed in landscape
orientation are printed on landscape-oriented paper.
[0111] In the embodiment of FIG. 14a, the image processor 1005 and
the print processor 1007 are shown to be distinct processors;
in-other embodiments, the tasks of these processors are performed
by a combined image and print processor.
[0112] FIG. 14b shows an alternative embodiment having the same
functional units 1001-1007 of the copier 1000 of FIG. 14a; however,
these units are not integrated in one and the same device. Rather,
a separate scanner 1003 and a separate printer 1006 are provided.
The data processing and data storing units, i.e. the memory 1004,
the image processor 1005 and the print processor 1007 (here called
"reproduction processor") may be part of a separate special-purpose
or multi-purpose computer, or may be integrated in the scanner 1003
and/or the printer 1006.
[0113] FIG. 14b also shows another reproducing device, a display
screen 1011. The display screen 1011 may replace, or may be used in
addition to, the printer 1006. When the screen 1011 is used to
reproduce the images, typically no print-direction tagging is
applied.
[0114] FIGS. 14c and 14d illustrate embodiments of a display screen
(FIG. 14c) and a printer (FIG. 14d) in which the image processor
1005 and the reproduction processor 1006 are integrated in the
display screen 1011 and the printer 1006, respectively.
Consequently, such screens and printers perform the described
image-quality improvements in a stand-alone manner and can
therefore be coupled to usual image-data sources, such as a usual
multi-purpose computer, which need not be specifically arranged to
provide, or even have awareness of, the image-quality-improving
measures applied.
[0115] FIGS. 15 and 16 are high-level functional diagrams of the
image processor 1005 and the reproduction or print processor 1007
of FIG. 14. According to the representations of FIGS. 15 and 16,
the image processor 1005 and the reproduction processor 1007 are
subdivided into several components. However, it should be noted
that this subdivision is only functional and does not necessarily
imply a corresponding structural division. Typically, the
functional components shown represent functionalities of one or
more computer programs which do not necessarily have a component
structure, as the one shown in FIGS. 15 and 16. The functional
components shown can, of course, be merged with other functional
components or can be made of several distinct functional
sub-components.
[0116] According to FIG. 15, the image processor 1005 has an input
to receive bitmap-input images and an output to supply transformed
and/or tagged bitmap images to downstream reproduction processor
1007. A text finder 1100 is arranged to identify text zones within
the bitmap-input image, by means of a zoning-analysis algorithm. A
color determiner 1101 is arranged to determine, for each text item
(e.g. character) in the text zones found, the average color of the
pixels belonging to the text item. In some embodiments, the
definition of which pixels belong to a text item is based on OCR.
Based on the average color found, the color determiner 1001 is
further arranged to determine whether the pixels of a character are
close to a primary color or black so as to be snapped to the
primary color or black. A text-size determiner 1102 is arranged to
determine the sizes of the text items (e.g. characters) in the text
zones, for example based on OCR. A text-orientation determiner 1103
is arranged to determine the orientations of the individual text
items (e.g. characters) in the text zones for a page, and, based on
that, to determine the main text orientation and main reading
direction. A color transformer 1104 is arranged, based on the
results obtained by the color determiner 1101, to transform, in the
input-bitmap image, the color of pixels of characters to be snapped
to the respective primary color or black. Alternatively, a color
tagger 1105 is provided; it is arranged, based on the results
obtained by the color determiner 1101, to tag the pixels of
characters to be snapped so as to indicate that these pixels are to
be reproduced in the respective primary color or black. A
small-text tagger 1106 is arranged, based on the results obtained
by the text-size determiner 1102, to tag pixels or pixel regions of
small characters so as to indicate that these pixels, or pixel
regions, are to be reproduced with a higher spatial resolution.
Finally, a text-orientation tagger 1107 is arranged, based on the
determined main-reading directions of the individual pages, to tag
the pages so as to indicate whether they are to be printed in
portrait or landscape format, so as to assure that the print
direction for each page is perpendicular to the page's main reading
direction.
[0117] According to FIG. 16, the reproduction (or print) processor
1107 has an input to receive tagged images and an output to
directly control the image reproduction, e.g. to direct the print
head of an ink-jet printing device. A tagged-color selector 1110 is
arranged to cause bitmaps in which certain bits or bit regions are
color-tagged to be reproduced in the primary color, or black
indicated by the color tag. A print-mask processor 1111 is
arranged, on the basis of small-text tags assigned to the input
image, to prepare a print mask which causes the tagged
small-character regions to be reproduced with a higher spatial
resolution than the other text regions. A page-orientation turner
and print-media-tray selector 1112 is arranged, based on
text-orientation tags associated with pages of the input image, to
turn the image to be printed and select the appropriate print-media
tray (i.e. either the portrait tray or the landscape tray) so as to
assure that the print direction is perpendicular to the page's main
reading direction.
[0118] The preferred embodiments enable images containing text to
be reproduced with an improved text image quality and/or higher
throughput.
[0119] All publications and existing systems mentioned in this
specification are herein incorporated by reference.
[0120] Although certain methods and products constructed in
accordance with the teachings of the invention have been described
herein, the scope of coverage of this patent is not limited
thereto. On the contrary, this patent covers all embodiments of the
teachings of the invention fairly falling within the scope of the
appended claims either literally or under the doctrine of
equivalents.
* * * * *